COACH Clinician Forum 2015

Transcription

1 COACH Clinician Forum 2015 Making Connections: Trends & Tempos in Clinical Informatics & Professional Practice Big Data, Cognitive Computing and the Impact on Clinical Decision Support May 31 st, 2015 V IBM Corporation

2 Three IT shifts are candidates to enable healthcare transformation - enabling a Clinic without Walls a future based on extended reach, vertical integration and smarter care. Data is the new basis of strategic advantage Cloud is the path to new business models Engagement requires a systematic approach Data is becoming the world s new natural resource, transforming industries and professions. The emergence of cloud is transforming IT and business processes into digital services. Mobile is transforming individual engagement increasing ability to deliver value via extended reach. 2

3 Create a new vision for health systems Systems of Insight and Engagement are mandatory to complement our investments in Systems of Record.. Create care coordination models Appropriate Standardized Process Create Consumer Advocates Systems of engagement Systems of insight Systems of record Manage risk Improve IT economics Optimize operations patient flow 3

4 Systems of Insight analytic capabilities and health data scientists will redefine Clinical Decision Support. EMR Summary of treatment EMR Summary of Risks Checking to prevent errors doing it right Suggestions / Actions doing the right things EMR Summary of what was done EMR summary of what is missing Retrospective Knowledge Real-time best evidence Knowledge enabled by Referral Democratization of Knowledge 4

5 Agenda Shifting Sands Healthcare and Big Data Knowledge and Data-driven Analytics Watson and Emergence of Cognitive Computing Discussion 5

6 Global trends are forcing integration of health care and social services delivery and segmentation of populations based on social determinants of health. Aging Population 66% 2 million aged over 65 in UK will lack informal care from adult offspring by 2030 a 66% increase from 1.2 million in 2012 Increasing Costs $7 trillion+ The cost for health and social programs worldwide health is 50% of Ontario s budget squeezing out other programs Chronic Disease 1 out of 4 1 in 4 Canadians and 2 of 3 over the age of 65 have multiple chronic conditions. 5% of Ontario patients consume 66% of health costs 6

7 Costs Health ecosystems are responding to be more proactive, driving the need to identify patients at risk, and intervene in time. Reduce costs and improve quality with with proactive care Shift Resources Rising costs: Aging populations Chronic disease Complex conditions Wellness Costs increase along the continuum of care Diagnosis and Disease Early Intervention Maintenance Late Stage/ Co-Morbidity Mgmt. Evidencebased care protocols Promote healthy behaviors Proactive early detection Best practice care protocols Manage Closer to Home Care coordination for complex conditions 7

8 Path Toward Personalized Medicine Green, ED et al (2011). Charting a course for genomic medicine from base pairs to bedside. Nature 470: Change in personalized healthcare investment from 2005 to Biopharmaceutical companies investing in personalized healthcare research in Prominent personalized medicine treatments & diagnostics available 2 75% 94% 13 in in Tufts Center for the Study of Drug Development, 2010; 2 Personalized Medicine Coalition,

9 Decreasing Cost of Genome Sequencing 9

10 Health data from new sources is exploding and forcing reconsideration of investments that are aligned with future potential. Exogenous determinants 60% Impact on individual health from behavior and environment 1,100 Terabytes Generated per lifetime Volume, Variety, Velocity, Veracity Fitbits, Home Monitoring Systems, Educational records, Employment Status, Social Security accounts, Mental Health records, Caseworker files and more Genomic determinants 30% Impact on individual health 6 TB Generated per lifetime Volume Clinical determinants 10% Impact on individual health 0.4 TB Generated per lifetime Variety 10 Source: "The Relative Contribution of Multiple Determinants to Health Outcomes", Lauren McGover et al., Health Affairs, 33, no.2 (2014)

11 For Healthcare, harnessing Data as a new natural resource to strategic advantage means managing the Four Vs. Volume Velocity Variety Veracity Data at Rest Data in Motion Data in Many Forms Data in Doubt Terabytes to exabytes of existing data to process Streaming data, 1000 s per millisecond with seconds to respond Structured, unstructured, text, multimedia Data inconsistency & incompleteness, ambiguities, latency, deception, model approximations 11

12 Agenda Shifting Sands Healthcare and Big Data Knowledge and Data-driven Analytics Watson and Emergence of Cognitive Computing Discussion 12

13 Systems of Insight have the potential to redefine clinical decision support in the context of both Knowledge-driven and Data-driven Analytics. Knowledge-Driven Data-Driven Published Knowledge Observational Data Journals Books Guidelines Discovery Advisor Oncology Advisor Engagement Advisor Clinical Trials Matching EMR Assist Genomics Advisor Longitudinal records Claims, Rx, Labs Genomics Exogenous Patient similarity Predictive modeling Real World Evidence Genomics Visual analytics 13

14 Insights also means moving from retrospective to predictive e.g. monitoring of premature infants - predicting onset 24 hrs in advance. Research Project between University of Ontario Institute of Technology (UOIT) & IBM Research - Monitors infants in the Hospital for Sick Children NICU Manages 100 s data values per second applies context and evidence-based rules Measures trends in multiple readings and based on combination of subtle changes, predicts adverse event 24 hours before onset. 14

15 Insight means leveraging clinical notes - extraction of data from nonstructured text to structured factors via Watson NLP. Watson s Text Analytics Natural Language Processing Manages abbreviation, negation, ambiguous phrases 55% = LVEF Patient does not show signs = Negative Symptom + Pre-built Watson Annotators for Care Management Infer meaning from non-contextual content Cut back from two packs to one per day = Smoker Identify, normalize, and code medical and social facts in unstructured content including: Diagnosis, procedure, allergy, medication, labs, lifestyle A 42-year old male presents for a physical. He lives alone and recently cut back from 2 packs to 1 pack per day. He recently had a right hemicolectomy invasive grade 2 (of 4) adenocarcinoma in the ilocecal valve was found and excised. At the same time he had an appendectomy. The appendix showed no diagnostic abnormality. 15 Patient Age: 42 Gender: Male Hx Procedure Hx Procedure Smoker: Yes Living Arrangements: Alone hemicolectomy diagnosis: invasive adenocarcinoma anatomical site: ileocecal valve grade: 2 (of 4) appendectomy diagnosis: normal anatomical site: appendix 15

16 For example, using both structured and non-structured data, a predictive model identifies patients at risk for developing CHF patients identified for at-risk care management program Improves patient satisfaction and delays or prevents chronic disease onset Reduces Costs of care management through prioritization of resources Business Challenge: The problem is identifying those patients at risk before the onset of the disease and before diagnosis so that preventative steps can be taken. The Smarter Solution: IBM built and deployed a predictive model that achieved an accuracy score of 85% in its ability to identify those patients at risk for developing CHF in a 1 year timeframe. The model leveraged structured data from the EPIC EMR system and also incorporated unstructured clinical notes that helped the model better identify patients with the condition, increasing its accuracy 16

17 Systems of Insight and Data-driven Healthcare Analytics means Building Capabilities. Data Driven Healthcare Analytics Framework Care Mgmt. & Coordination Practice Management Real World Evidence Point of Care Decision Support IBM s Track Record in Healthcare Analytics Over 50 publications in 4 years Over 30 patents filed Nominated for best paper: AMIA 2010, AMIA Summit 2014 Received NIH grant for CHF prediction ( ) Data Patients Providers Payment Predictive Modeling ANALYTICS Risk Stratification Clinical Pathway Mining Utilization Analysis Translational Medicine Disease Progression Modeling Patient Similarity VISUAL ANALYTICS Visualization for Patient Clusters Visualization for Patient Evolutions 17

18 Large Scale Patient Similarity Analysis? x 1 Q x 2 Q x N Q Similarity Analysis? Clinically similar to Query patient Patient similarity assessment in clinical factor/feature space Patient population x 1 1 x 2 1 x N 1 x 2 1 x 2 2,,, x N 2 x 1 K x 2 K x N K Best Treatment=? Prognosis=? Diagnosis=? Outcomes Analysis Treatment Comparison Disease Progression Often > 10,000 dimensions Challenge: identify the measure of clinical similarity between patients Approach: machine learning algorithms to automatically learn the best metric from observational data and labels provided by experts or derived from data 18

19 IBM Research has evolved a Decision Support Tool for Clinical Care visualizes what happened to patients just like yours Key capabilities Uses Patient Similarity Analytics to find clinically similar patients. Clinical histories database Characteristics of current patient Extracts historical event trails and patient characteristics relevant to the condition targeted Provides a Visual Summary of the evolution of clinical pathways of the similar patients, based on relevant to events Relates outcomes to pathways, to help identifies most desirable and most problematic pathways to inform clinical decisions Extract historical patient event trails and patient characteristics CareFlow Visualization Patient Similarity Analytics 19

20 IBM Research has evolved a Decision Support Tool for Clinical Care visualizes what happened to patients just like yours 20

21 Transforming Skin Cancer Detection using Data-Driven Learning Today Transformation 1 (Current) Transformation 2 (Future) Diagnosis Computer Assisted Diagnosis Diagnosis and Continuous Learning Manual Inspection ABCDE Asymmetry Borders Color Diameter Evolution Computer Analysis Ensemble Classification Segmentation and Visual Feature Extraction Deep Learning Convolutional Neural Network (CNN) Discrimination and Feature Learning Unknown Images Unknown Images Unknown Images Inaccurate Diagnoses from Limited Samples and Modalities Subjective Evaluation Higher Accuracy (1K s training) Improved Consistency More Contextual Data (modalities) Manual Data-Driven Highly Accurate (1M s training) Continuous Learning from Data High Density Samples (space/time) 21

22 Using Systems of Insight to know your customers segment patients similar to Catalunya, Spain who have risk profiled based on 8 Clinical Risk Groups. 22

23 Agenda Shifting Sands Healthcare and Analytics Knowledge and Data-driven Analytics Watson and Emergence of Cognitive Computing Discussion 23

24 Knowledge-driven Analytics and emergence of Cognitive Computing Learning Systems - to address the challenges of healthcare Watson focuses on Knowledge Management EMR investments focus on record management 800,000 publications per year catalogued by the US Library of Medicine 24

25 Starting in 2006, our investments have led to Watson for Healthcare - integration of knowledge with data to drive better decisions and outcomes. 1 Understands natural language and human speech 2 Generates and evaluates hypothesis for better outcomes 99% 60% 10% 3 Adapts and Learns from user selections and responses 25

26 Watson s Cognitive Computing capabilities allows you to Ask, Discover, and Decide. ASK Watson Engagement Advisor to transform interactions and experiences with consumers and patients Emerging Technologies Watson Paths for medical learning DISCOVER Watson Discovery Advisor Watson Analyticsl Watson Explorer to consolidate and visualize information Emerging Technologies Watson EMR Assistant Identify critical attributes of a patient case and provide easy-to-consume summaries DECIDE Watson Oncology to assist in identifying individualized treatment plans and clinical trials.. 26

27 IBM Watson for Oncology Trained by Memorial Sloan Kettering Business challenge: Ability to assess quickly the best treatments for an individual patient based on latest evidence and clinical guidelines Watson solution: A tool to assist physicians make personalized treatment decisions Analyzes patient data against thousands of historical cases and trained through thousands of Memorial Sloan Kettering MD and analyst hours Suggestions to help inform oncologists decisions based on over 290 medical journals, over 200 textbooks, and 12M pages of text Evolves with the fast-changing field Currently supports first line treatment (Breast, Lung, Colorectal cancers) 27

28 Watson Oncology Overview 2 Search a corpus of Use those attributes to find evidence data to find candidate treatment options supporting evidence for as determined by each option consulting NCCN Guidelines Guidelines 3 61 y/o woman s/p mastectomy is here to discuss treatment options for a recently diagnosed 4.2 cm grade 2 infiltrating ductal carcinoma Patient Case Key Case Attributes Candidate Treatment Options Watson Oncology Supporting Evidence Inclusion / exclusion criteria Co morbidities Contraindications FDA risk factors MSK preferred treatments Other guidelines Published literature - studies, reports, opinions from Text Books, Journals, Manuals, etc. 1 Evidence Extract key attributes from a patient s case Prioritized Treatment Options + Evidence Profile 4 Use Watson s analytic algorithms to prioritize treatment options based on best evidence. 28

29 29

30 30

31 31

32 Clinical Trial Matching is being used by The Mayo Clinic to increase participation rates in clinical trials. Business challenge: Clinicians have no easy way to search across eligibility criteria of relevant clinical trials for their patient; 30% of sites for clinical trials fail in enrolling even a single patient Watson solution: Use patient data to instantly check eligibility across all relevant clinical trials Initial focus is oncology, Watson is trained in breast, lung and colorectal cancer, with other tumor types to follow 32 32

33 Department of Veterans Affairs will assess Watson technology in pilot study that could benefit 8.3 million veterans requiring care each year The amount of medical data doubles every three years Size and complexity associated with patient data in EMRs is overwhelming The potential of EMRs has not been realized given the discrepancies of how the data is recorded, collected and organized across healthcare systems Success Value Challenges Watson will make it possible for VHA physicians to interact with medical data in natural language, process millions of pages of patient information and medical literature to uncover patterns and insights, and learn from each interaction. By sifting through reams of clinical data, Watson is able to distill evidence and knowledge within seconds. During the pilot, Watson will base clinical decisions on realistic simulations of patient encounters pre-visit, visit and post-visit situations. Physicians can save valuable time finding the right information needed to care for their patients with Watson technology A tool that can help a clinician quickly collect, combine and present information will allow them to spend more time listening and interacting with the Veteran. Carolyn Clancy - Interim Undersecretary for Health - VA 33 33

34 New York Genome Center and IBM Watson launched an initiative to accelerate a new era of genomic medicine Success Value Challenges As the cost of Next Generation Sequencing decreases, there will be an increase in tumor genome sequencing resulting in massive quantities of genetic data to analyze It can take on an average from 4-6 weeks to analyze and interpret genetic data manually Complexity of matching genetic mutations of individual s tumor with molecular targeted therapies using multiple data sources "With this [genomics] knowledge, doctors will be able to attack cancer and other devastating diseases with treatments that are tailored to the patient s and disease s own DNA profiles. This is a major transformation that can help improve the lives of millions of patients around the world. Dr. John E. Kelly, Senior Vice President, Solutions Portfolio & Research Secured Beta testing relationships with 13 Cancer and Academic medical centers Applying the cognitive computing power of Watson is going to revolutionize genomics and accelerate the opportunity to improve outcomes for patients with deadly diseases by providing personalized treatment. Robert Darnell, M.D., Ph.D., CEO, President and Scientific Director of the New York Genome Center Source:

35 Continued growth in the amount and complexity of medical knowledge leverage Watson to scale expertise. Clinical trials Ingest Learn Test Reference Genomes Mutation Protein Pathways Genomic Data Clinical Treatments Scientific Literature Pharmaceutical Reports Patient Reports Chemical Patents Medline Dysfunctional Proteins &Targeted Treatments 35

36 Recap: Systems of Insight Knowledge-driven and Data-driven shifting clinical decision support. EMR summary of Risks Suggestions / Actions doing the right things EMR summary of what is missing Real-time best evidence Democratization of Knowledge Cognitive Computing 36

37 Discussion 37

38 Together with our Cloud investments, we were positioned to enable IBM Watson Health Human Services 1. Data - HIPAA-enabled, standards-based, massively scalable, open repository of data on all dimensions of health 2. Insights as a Service providing knowledge and actionable information through advanced analytics and cognitive capabilities 3. Solutions from IBM and ecosystem partners improves the overall experience and increases the quality of outcomes -- where it matters and when it matters Health Smarter Care Social Programs 38

39 IBM Watson Health consists of Acquisitions Partnerships IBM organic innovation IBM Cúram Smarter Care Social Programs IBM Research Assets Data and solutions Massive big data management Solutions for population health and patient engagement Insights Population health Patient engagement Insights and solutions Care management Human services delivery 39